San Francisco artificial intelligence lab Radical Numerics announced it administered a custom AI-designed bacteriophage to a human patient with a severe bacterial infection under an FDA-authorized compassionate use pathway.
The fact that this phage could be administered to a patient is one of the first examples of what this approach can enable
, Eric Nguyen said, according to BigGo Finance.
An artificial intelligence research lab in California has administered a custom AI-designed virus to a human patient for the first time (according to Yahoo reporting). The treatment, created to target an antibiotic-resistant bacterial infection, makes use of bacteriophages—viruses that specifically infect and destroy bacteria. The milestone centers on Omnii-1, a multimodal genome language model launched by Radical Numerics, an AI research lab founded by members of the team behind Evo and Evo 2.
Across the globe, antimicrobial resistance accounted for roughly 1.27 million direct fatalities in 2019 out of an estimated 4.95 million linked deaths, while forecasts indicate that superbugs resistant to antibiotics could trigger nearly 40 million deaths by the year 2050.
Laboratory Testing of Omnii-1 Phage Variants
Built by the team behind Evo and Evo 2, Omnii-1 is a biological AI model that learns across DNA, RNA, proteins, and other molecular data. The startup based its work on the T7 bacteriophage, which infects Escherichia coli. Researchers synthesized the proposed DNA sequences and produced modified viruses in the lab.
Introducing Omnii-1, our most powerful genome language model, and a major step toward general biological intelligenceIn a world first, Omnii-1 designed the first AI-generated bacteriophages to be administered to a human patient with an antibiotic-resistant infection under… pic.twitter.com/3Bc1sjr9Qt
Before clinical administration, the lab tested the resulting phage variants against bacteria isolated from the patient. As The Register noted, a cocktail containing the Omnii-1-designed phage variants eliminated the patient’s antibiotic-resistant isolate in in vitro testing.
The cocktail containing Omnii-1-designed phage variants eliminated the patient’s antibiotic-resistant isolate in vitro
, the company explained in a technical report.
Investigators at Stanford University published findings in August 2026 that lay the groundwork for this effort. Investigators at the Arc Institute also published findings in August 2026 that lay the groundwork for this effort.
They demonstrated that viral genomes generated by artificial intelligence could successfully yield working bacteriophages within a laboratory setting. Functional variants made up 69.9% of the total proposals generated by Omnii-1, contrasting sharply with a success rate of only 4.8% observed among variants created through random modifications of a comparable magnitude. Several of the tested variants outperformed the original T7 phage.
Food and Drug Administration Compassionate Use Authorization
Following successful laboratory testing, the treatment was handed over to attending physicians. DongA Science reported that the startup carried out the treatment under authorization from the US Food and Drug Administration. The therapy was administered at a clinical site under physician-led compassionate use, an FDA pathway allowing physicians to request access to investigational treatments for patients with serious or life-threatening conditions who have no comparable treatment alternatives.
Eric Nguyen stated that the company thinks it can design and build phages targeted to individual bacterial infections in a matter of days.
Radical Numerics Limits Disclosure of Patient and Doctor Details
Radical Numerics has not disclosed specific details regarding the patient’s clinical progress, condition, or treatment outcomes. Citing patient privacy and the proprietary nature of a treatment it hopes to monetize, the company declined to share information about the participating doctors or the exact authorization date.
We take our responsibility to protect patient confidentiality and maintain the trust of our clinical partners extremely seriously
, Nguyen stated, adding that the firm intends to publish its scientific findings through peer-reviewed research.
Specialists have responded to the announcement with a mixture of enthusiasm and caution. Michael Baym, a microbiologist at Harvard University who was not involved in the research, offered a measured reaction to ForkLog and other outlets.
If it works, it will be amazing.
Michael Baym, microbiologist at Harvard University
Baym added that he would refrain from making broader assessments until the treatment is completed and independent observers have an opportunity to evaluate how novel the resulting viruses actually were. The company itself cautioned that observations from a single patient do not establish general safety or effectiveness, noting that additional peer-reviewed scrutiny remains necessary.
Readers concluding their review of these events should consult qualified medical professionals or infectious disease specialists regarding any personal health concerns, antibiotic resistance questions, or treatment decisions, as single-patient experimental reports do not substitute for formal clinical guidance.